How well do we know the electromagnetic form factors of the proton?

Several experiments have extracted proton electromagnetic form factors from elastic cross section measurements using the Rosenbluth technique. Global analyses of these measurements indicate approximate scaling of the electric and magnetic form factors (mu_p*G_E/G_M = 1), in contrast to recent polarization transfer measurements from Jefferson Lab. We present here a global reanalysis of the cross section data aimed at understanding the disagreement between the Rosenbluth extraction and the polarization transfer data. We find that the individual cross section measurements are self-consistent, and that the new global analysis yields results that are still inconsistent with polarization measurements. This discrepancy indicates a fundamental problem in one of the two techniques, or a significant error in polarization transfer or cross section measurements. An error in the polarization data would imply a large error in the extracted electric form factor, while an error in the cross sections implies an uncertainty in the extracted form factors, even if the form factor ratiois measured exactly.